Impact of the LMDZ atmospheric grid configuration on the climate and sensitivity of the IPSL-CM5A coupled model

The IPSL-CM5A climate model was used to perform a large number of control, historical and climate change simulations in the frame of CMIP5. The refined horizontal and vertical grid of the atmospheric component, LMDZ, constitutes a major difference compared to the previous IPSL-CM4 version used for CMIP3. From imposed-SST (Sea Surface Temperature) and coupled numerical experiments, we systematically analyze the impact of the horizontal and vertical grid resolution on the simulated climate. The refinement of the horizontal grid results in a systematic reduction of major biases in the mean tropospheric structures and SST. The mid-latitude jets, located too close to the equator with the coarsest grids, move poleward. This robust feature, is accompanied by a drying at mid-latitudes and a reduction of cold biases in mid-latitudes relative to the equator. The model was also extended to the stratosphere by increasing the number of layers on the vertical from 19 to 39 (15 in the stratosphere) and adding relevant parameterizations. The 39-layer version captures the dominant modes of the stratospheric variability and exhibits stratospheric sudden warmings. Changing either the vertical or horizontal resolution modifies the global energy balance in imposed-SST simulations by typically several W/m² which translates in the coupled atmosphere-ocean simulations into a different global-mean SST. The sensitivity is of about 1.2 K per 1 W/m² when varying the horizontal grid. A re-tuning of model parameters was thus required to restore this energy balance in the imposed-SST simulations and reduce the biases in the simulated mean surface temperature and, to some extent, latitudinal SST variations in the coupled experiments for the modern climate. The tuning hardly compensates, however, for robust biases of the coupled model. Despite the wide range of grid configurations explored and their significant impact on the present-day climate, the climate sensitivity remains essentially unchanged.     

Ocean expansion: The role of detachment faults

Detachment faulting at slow spreading ocean ridge axes is recognized as a major surface creation mechanism, yet the structural relationships of these faults with feeder-dykes of on-axis volcanoes remained unresolved. This summary study shows that surface creation leading to ocean widening is exclusively controlled by detachment faults, shallow tracers of tectonic stresses induced by the westward drift of plates. Volcanoes are fed by feeder-dykes following on-axis rotational detachment faults. Once formed volcanoes are dragged along the detachment whose footwall is made of mantle material sometimes hosting gabbro sills. Due to the faster drift of the uppermost lithospheric layer, the feeder-dykes are then intersected by active deeper low-angle detachments, become inactive and are replaced by new ones on-axis. Rooted vertically on either side of the plate boundary, the detachment flexing is all the earlier and more progressive as the faults are far from the axis, positioning gabbro as sills at shallow level within a deformed mantle interspersed with cataclasite horizons. Correlations between shallow and deep lithospheric processes are then clarified.     

Acceleration of thermodynamic computations in fluid flow applications

Reservoir simulators model the highly nonlinear partial differential equations that represent flows in heterogeneous porous media. The system is made up of conservation equations for each thermodynamic species, flash equilibrium equations and some constraints. With advances in Field Development Planning (FDP) strategies, clients need to model highly complex Improved Oil Recovery processes such as gas re-injection and CO2 injection, which requires multi-component simulation models. The operating range of these simulation models is usually around the mixture critical point and this can be very difficult to simulate due to phase mislabeling and poor nonlinear convergence. We present a Machine Learning (ML) based approach that significantly accelerates such simulation models. One of the most important physical parameters required in order to simulate complex fluids in the subsurface is the critical temperature (T_crit). There are advanced iterative methods to compute the critical point such as the algorithm proposed by Heidemann and Khalil (AIChE J 26,769–799, 1980) but, because these methods are too expensive, they are usually replaced by cheaper and less accurate methods such as the Li-correlation (Reid and Sherwood 1966). In this work we use a ML workflow that is based on two interacting fully connected neural networks, one a classifier and the other a regressor, that are used to replace physical algorithms for single phase labelling and improve the convergence of the simulator. We generate real time compositional training data using a linear mixing rule between the injected and the in-situ fluid compositions that can exhibit temporal evolution. In many complicated scenarios, a physical critical temperature does not exist and the iterative sequence fails to converge. We train the classifier to identify, a-priori, if a sequence of iterations will diverge. The regressor is then trained to predict an accurate value of T_crit. A framework is developed inside the simulator based on TensorFlow that aids real time machine learning applications. The training data is generated within the simulator at the beginning of the simulation run and the ML models are trained on this data while the simulator is running. All the run-times presented in this paper include the time taken to generate the training data and train the models. Applying this ML workflow to real field gas re-injection cases suffering from severe convergence issues has resulted in a 10-fold reduction of the nonlinear iterations in the examples shown in this paper, with the overall run time reduced 2- to 10-fold, thus making complex FDP workflows several times faster. Such models are usually run many times in history matching and optimization workflows, which results in compounded computational savings. The workflow also results in more accurate prediction of the oil in place due to better single phase labelling.